The $202 everyone quotes? Three out of four advertisers paid more.
So the bar the market quotes is already a good result. Here is where you actually sit.
The number your team is measured against was never the middle of anything: it is cheaper than three-quarters of advertisers managed, so most teams are chasing a bar the market never cleared. This page gives you the honest version of your own numbers. Where your cost per lead sits against what everyone else actually paid, what your leads are worth once they turn into customers, and what an outside discount range does to the return your ad platforms report. For scale: across these advertisers, ads returned 0.56x in the first year, at $58,887 per customer and a 21.2% close rate. That is credit we can trace, not proof the ads caused it.
Three inputs, three answers you can put in Monday’s report: where your cost per lead lands against what every other advertiser actually paid, not against an average weighted by who spent most, your customers per 1,000 leads next to built audiences (3.56) and native targeting (2.6), and a discounted range, never one tidy number, for the return your platforms report. No email gate, nothing stored, nothing sent anywhere.
2025 · 153 advertisers · $57.6M analysed · customer numbers from the $29.4M lead-gen slice (127 advertisers, 154,000 leads matched to a CRM) · Methodology
Executive summary
- The gap: the benchmark the market quotes is already a good result. Three out of four advertisers paid more than the $202 everyone cites, so your team is being marked against a bar most of the market never cleared.
- The board move: retire the single blended cost per lead and take the six numbers below to the next review. Cost per lead, customers per 1,000 leads, close rate, cost per customer, payback you can trace, and average deal won. The Board page prints them on one sheet.
- The caveat that travels with it: this is credit we can trace, not proof the ads caused it, not profit and not lifetime value. Read it as a floor and a starting point.
Playbook
- The play: check your cost per lead against what everyone else actually paid before you defend it or apologise for it. The $202 everyone quotes for LinkedIn is cheaper than three out of four advertisers managed. Most teams are quoting a number they never hit.
- The setup: put last quarter’s leads and won customers into the second tab and compare yourself with built audiences at 3.56 customers per 1,000 leads and native targeting at 2.6.
- The guard: discount the return your platforms report as a range, and say out loud that the range comes from other people’s research, not from our data. One tidy number here is a fabrication with a decimal point on it.
Put your own numbers in
Three questions, answered off the published data: where you sit against everyone else, what your leads turn into, and what the return your platforms report looks like after an outside discount range is applied. Nothing is gated, nothing is stored, nothing is estimated, where a number failed our publication rules, this tool says so instead of guessing.
How does your cost per lead compare with what everyone else paid?
Pick a channel and type in what you pay for a lead. We place it against what every advertiser in that channel actually paid, one advertiser, one data point, and hand back the stretch of the range you land in. The $202 everyone quotes for LinkedIn is cheaper than three out of four advertisers managed: only about a quarter of them beat it, and those few do not account for most of the spending.
What do 1,000 of your leads become?
The reporting change this data argues for: judge yourself on customers per 1,000 leads and cost per customer, not on cost per lead. Metadata-built audiences cost 20% more per lead ($217 against native targeting’s $181) and produced 37% more customers per 1,000 leads, 3.56 against 2.6, at a 13% lower cost per customer ($60,896 against $69,705). The caveat rides in the same breath: native traced back more revenue per dollar, 0.71x against 0.43x, entirely because its won deals were 1.9x bigger ($49,778 against $26,064). We have not worked out why those deals were bigger, so the claim here is better conversion, not more revenue.
Discount your dashboard
Type in the return your ad platforms report. We divide it by how much other people’s research says platforms like yours overstate, and hand back a range. That range is not ours and is not a Metadata measurement: it comes from 2026 studies by other researchers, who held ads back from some accounts to see what would have sold anyway, and who report the overstatement as a spread, not a single figure. Our own comparable number, counted on deals that actually closed in the CRM, is 0.56x, a floor, not a verdict.
The Board page
One sheet a director can forward to a CMO, or a CMO can drop into a review: your numbers, the six that matter, what each one means, the rule that stops any one advertiser swinging a number, and how to cite it. Cmd-P / Ctrl-P prints this section on its own.
Metadata 2026 B2B Ad Spend Benchmark, self-benchmark sheet, metadata.io/benchmark-report-2026/insights/reality-gap
Paid media reality check
2025 · 153 advertisers · $57.6M analysed · customer numbers from the $29.4M lead-gen sliceYour numbers
- Fill the calculator above and this page fills itself.
Where you land against everyone else
,
The six numbers that matter
| What you are measuring | You | Built audiences (104 advertisers) | Native targeting (72 advertisers) | All advertisers |
|---|---|---|---|---|
| Cost per lead | — | $217 | $181 | see the range |
| Customers per 1,000 leads | — | 3.56 | 2.6 | — |
| Close rate, deals that actually closed or died | — | 19.2% | 17.4% | 21.2% |
| Cost per customer | — | $60,896 | $69,705 | $58,887 |
| Payback we can trace | — | 0.43x | 0.71x | 0.56x |
| Average won deal | — | $26,064 | $49,778 | — |
What these mean. Payback we can trace = revenue from deals that actually closed, with the credit split evenly across the ads that touched them, divided by what was spent on ads in the same period. It is credit we can trace, not proof the ads caused it, not profit, not lifetime value. Close rate = we counted only deals that actually closed or actually died; deals still open are left out of the sum and reported on their own. Cost per customer = ad spend divided by the customers that even credit split assigns to the ads. Customers per 1,000 leads = those same customers ÷ leads × 1,000.
The period. The 12 months of calendar 2025. The alternatives our reviewers split over, 6 to 18 months matched to the sales cycle, 180 to 365 days, and 90 days with a plain warning that it undercounts. Are printed in the methodology rather than averaged away.
No one advertiser may swing a number. Any number where a single advertiser holds more than 40% of the results is held back, and we wrote that rule down before we saw the numbers. Retargeting’s cost-per-customer numbers broke it, so this report publishes no retargeting customers per 1,000 leads, no cost per customer and no close rate. Cost and click numbers also need 5 advertisers and $50,000 of spend in a group before we publish them.
The discount range. The range in the third tab is other people’s 2026 research applied to the number your platforms gave you. It is not a Metadata measurement, and it is a range because the research reports a range.
Cite as: Metadata 2026 B2B Ad Spend Benchmark (n=153 advertisers, $57.6M, 2025), metadata.io/benchmark-report-2026
How do we know?
Every number this tool gives back is either your own arithmetic or a number we published. Where a number did not survive our publication rules, the tool says so instead of filling the gap.
- What everyone else paid is real, counted one advertiser at a time: 88 LinkedIn advertisers, 40 Facebook, 28 Instagram, 34 Google Ads, published at five marks. The cheapest tenth, the cheapest quarter, the middle, where the priciest quarter starts, and the top 10%. We place you between two of those marks. We never print an exact position the data cannot support.
- Two ways of counting, both shown: one advertiser at a time, and weighted by how much each advertiser spent. The famous $202 LinkedIn figure is the spend-weighted one, and three out of four advertisers paid more than it, which is why “we are at benchmark” is a weaker statement than it sounds.
- What a customer number has to clear: at least 8 advertisers, at least 3 deals actually won, and no advertiser holding more than 40% of the result. Cost and click numbers need 5 advertisers, $50,000 of spend and no advertiser above 50% of the group’s spend.
- Ours versus theirs: what advertisers paid, and what their leads became, is our own 2025 data. The discount range in the third tab is other people’s research, labelled as theirs everywhere it appears, and it carries no claim marker because it is not ours to certify.
- What this cannot tell you: this is credit we can trace, not proof the ads caused it. Nothing here proves cause, and a discounted range is a sanity check on scale, not a measurement of what your ads added. Proving that needs holdouts, deliberately withholding ads from some accounts, regions or weeks. Which this data does not have.
- What kind of data this is: things that happened together in 2025, across advertisers who use Metadata and can match their ads to their CRM. That skews the picture toward teams whose data is clean enough to measure this at all.
Full rules in the methodology, including the claims we refused to publish, with the reason each was killed and the number it cost us.
Cite as: Metadata 2026 B2B Ad Spend Benchmark (n=153 advertisers, $57.6M, 2025), metadata.io/benchmark-report-2026